The present invention relates to a computer system capable of improving a response performance to a request from a client.
A main trend of recent information systems is a Web-based business system (hereinafter called a “Web system”) constituted of a plurality of system apparatuses including network apparatuses, a Web server group, an AP server group and a DB server group.
A number of unspecified clients access a Web system via the Internet so that a large number of accesses are concentrated in some cases. If accesses are concentrated upon the Web system and responses to users are delayed, users become reluctant to access the Web system. An enterprise running the Web system may lose business chances.
In order to deal with this problem, load distribution has been introduced conventionally. For example, the load is distributed by additionally installing a new server group to the Web system to shorten response times to users.
It is generally known that this load distribution method by additional installation is performed in the unit of business application. Namely, load distribution is performed by additionally installing as one unit a server group installing a business program group which constitutes a business application.
This load distribution method by additional installation in the unit of business application is not necessary to change parameters regarding a dependency relation (configuration, substitution control and the like) between servers to be additionally installed and between business programs. This method can therefore be performed relatively easily without halting the Web system.
There occurs, however, a problem that a large amount of system resources (hardware) is required because a server group installed with a business program group having a small load is also used as an additional installation object.
Furthermore, in a case that a load is concentrated upon a particular Web content in the business program and this load concentration adversely affects another Web content processed in the same business program, even if the load distribution method by additional server installation is performed, a response time to a client accessing the other Web content is not improved so much in some cases.
If a server group to be additionally installed and used between a plurality of Web systems has a large capacity of business programs, an installation time of the business programs to servers is prolonged.
In an urgent case not expected by a running manager, such as upon abrupt occurrence of a high load on a Web system, the installation time may become a bottleneck.
As one example of load distribution techniques through additional installation, JP-A-2003-178040 discloses techniques of calculating and deciding the number of servers to be additionally installed in accordance with performance information on a server group constituting a Web system. According to the techniques disclosed in JP-A-2003-178040, additional installation is performed on a server unit basis, so that the problem of using a large amount of system resources can be mitigated more than load distribution through additional installation on a business application basis.
According to JP-A-2003-178040, however, if a load is applied to a particular Web content in a business program, there is a possibility that a response time to another Web content is not improved so much even if the load is distributed through additional installation.